Re: df shows 100% on a empty filesystem-

From: Prasad <p4cldba_at_gmail.com>
Date: Wed, 16 Jan 2008 17:55:14 -0800
Message-ID: <666b99c70801161755y754ef84fmbad58bc26737461a@mail.gmail.com>


Hi Krish,

Even though alert log shows ARCHIVE LOG STOP has been deprecated I still tried alter system ARCHIVE LOG STOP for an hour ( as I have enabled the log_archive_dest_state_3 pointing to a different filesystem) but it didnt help me ..

I do see lsof pointing to /ofs/archivelog01 but the problem is that there are multiple unix pid owned by oracle doing something on /ofs/archivelogs01 .

Thanks
-Prasad

On Jan 16, 2008 5:05 PM, <krish.hariharan_at_quasardb.com> wrote:

> We have seen this in our large environments where several hundred gb gets
> tied up. Removing the files will not work since you are only manipulating
> the directory entries and unless the inode is released the OS will not
> release the space and that will happen only when the process that has those
> file handles open, closes them or is terminated. We end up shutting down the
> database and on occasion reboot the server. When you move files across file
> system you are again copying the files and then removing the directory
> entries in the source file systems. The process that has the files open is
> not affected by the rename/removal of a file's directory entry.
>
>
>
> I have looked at these processes using the files by using fuser but I
> haven't found that generally useful since it tends to dump all the oracle
> processes. The other utility that I know of but haven't used is lsof which
> gives the list of open files.
>
>
>
> A long shot: One approach you may be able to take is to issue an archive
> log stop for a brief period (not sure if this will stop the archiver or not
> since the alert log says it is deprecated and the processes are still
> hanging around) and see if it posts/causes the archiver to release the file
> handles and allow the os the release the space (assuming that you have
> sufficiently large and a large number of redo log groups such that
> transaction processing is not arrested).
>
>
>
> Regards,
>
> -Krish
>
> Krish Hariharan
>
> President/Executive Architect, Quasar Database Technologies, LLC
>
> http://www.linkedin.com/in/quasardb
> ------------------------------
>
> *From:* oracle-l-bounce_at_freelists.org [mailto:
> oracle-l-bounce_at_freelists.org] *On Behalf Of *Prasad
> *Sent:* Wednesday, January 16, 2008 5:01 PM
> *To:* Mir M. Mirhashimali
> *Cc:* oracle-l
> *Subject:* Re: df shows 100% on a empty filesystem-
>
>
>
> Hi Mir,
>
> This db is on solaris 8 . it do not have a lost and found . The du -a
> shows me the correct picture but the problem is i can not take down the db
> or unmount the file system now.
>
> thanks
> -Prasad
>
> On Jan 16, 2008 3:51 PM, Mir M. Mirhashimali <mhyder_at_rice.edu> wrote:
>
> Prasad,
>
> I was in a similar situation recently. even after moving files from the
> directory it was showing 100%. check "lost and found" or ".snapshot" i
> am on redhat linux so not sure if this applies to you.
>
> Hope this Helps
> Mir
>
>
> Prasad wrote:
> > All,
> >
> > The database is 10gR2 db running on solaris 8 .. There was archiver
> > hung on archlogs01 filesystem . I moved all the archive logs to the
> > backup filesystem . however df still shows archlogs01 100% filled.
> >
> > du shows correctly
> >
> > du -a /ofs/archlogs01
> > 4 /ofs/archlogs01/oracle/PLLS
> > 6 /ofs/archlogs01/oracle
> > 10 /ofs/archlogs01
> >
> > gdf -h shows incorrect
> >
> > gdf -h | grep /ofs/archlogs01
> > /dev/dsk/c2t0d11s0 2.9G 2.9G 2.0k 100% /ofs/archlogs01
> >
> > fuser -c do not show me any active process running on /ofs/archlogs01
> > archive log stop also didnt help .
> >
> > Any suggestions!
> >
> > TIA
> > Prasad
> >
>
> > !DSPAM:59,478e896a200741201525439!
>
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Wed Jan 16 2008 - 19:55:14 CST

Original text of this message